APM Usage

APM (Agent Package Manager) is the open-source package manager for AI coding agents. It lets teams install, share, and govern reusable instructions, prompts, agents, skills, and MCP server configurations across projects.

When to activate

• User mentions apm or "Agent Package Manager"
• Questions about installing or managing AI-agent packages
• Setting up instructions, prompts, agents, skills, or chatmodes
• Configuring MCP or LSP servers through apm.yml
• Authentication for private repos (GitHub, ADO, GHES, Artifactory)
• Policy enforcement or apm audit
• Package authoring or publishing
• Compiling agent context (apm compile)
• Troubleshooting apm errors

Key rules

• Commit these files: apm.yml, apm.lock.yaml, .apm/, deployed harness directories including .grok/, and AGENTS.md
• Never commit: apm_modules/ (add to .gitignore)
• Team sync: after git clone, run apm install to restore dependencies
• Update deps: apm install --update refreshes to latest refs
• Pin versions: use tags (#v1.0.0) in production, branches for development
• Prefer ASCII: keep CLI output and scripts to printable ASCII for portability; package content itself may be UTF-8

Reference

For detailed guidance, see the following resources:

• Installation -- install and update APM
• Workflow -- core workflow, apm.yml format, what to commit
• Commands -- full CLI command reference
• Dependencies -- all dependency formats, including MCP and LSP servers, and version pinning
• Authentication -- token setup for private repos
• Governance -- policy engine and audit checks
• Package Authoring -- creating APM packages
• Troubleshooting -- common errors and fixes